Add query execution timeout for runaway queries

Impala has the ability to timeout idle queries via the global flag -idle_query_timeout and the query option 'query_timeout_s' (which unfortunately doesn't include 'idle'), but it would be useful to also have a timeout that cancels queries after some amount of execution regardless of activity. This would be helpful to ensure some workloads don't have 'runaway' queries (in addition to mem limits), and should probably be capable of being set on a per-pool basis.

This shouldn't be too hard to add once we fix IMPALA-1575.